Re-raise exceptions from on_data handler

[martinv13]

Thank you for this very useful package.

Is your feature request related to a problem? Please describe.

I am using the on_data handler to process decrypted data. When an exception is raised within it, decrypt_file call prints it to the console but does not raise.

In the result returned by decrypt_file I see no clue that this exception was raised.

Describe the solution you'd like

I would prefer that decrypt_file re-raise the exception instead of printing it. Or alternatively that returncode were changed to non-zero. Or another possibility would be simply to warn about this behavior in the doc as I suppose most users would expect decrypt_file to re-raise in the first place (the example shown in the doc says that it does not handle errors but could be further specified).

Describe alternatives you've considered

I could circumvent this issue by handling all exceptions within the callable object passed to on_data, setting a flag when an exception is raised and read this flag afterwards to raise outside decrypt_file call but it does not look very intuitive.

Thanks!

[vsajip]

The on_data handling happens across several I/O operations, not just decryption, at a lower level than e.g. decrypt_file - so it would be hard to provide the solution you'd like. The exception is actually raised on a different thread, and printing could perhaps be avoided by swallowing an exception, but that doesn't seem ideal either. I don't mind adding suggestions for error handling in on_data to the documentation.

[martinv13]

Thanks for your answer, I didn't realize that it was used for other cases too.

Indeed adding something along those lines in the doc would be helpful:

Beware that decrypt_file will not re-raise exceptions thrown by the on_data handler, but rather print them to the console. As a consequence you should make sure that you handle them within the on_data handler.

And change one of the examples like this for instance:

class Collector:
        def __init__(self, fn):
            self.out = open(fn, 'wb')
            self.exception = None

        def __call__(self, chunk):
            try:
                self.out.write(chunk)
            except Exception as e:
                self.exception = e
            finally:
                if not chunk:
                    self.out.close()
            return False  # Tell python-gnupg not to buffer the chunk

I can do a quick PR if you'd like, but maybe you prefer writing this a bit differently?

Thanks
Martin

[vsajip]

I've already started on something to address this, I'll push it shortly - just quite busy this weekend. My approach is a little different (allowing on_data to just be a function) - it captures the exception in a low-level I/O handling routine and returns it in an on_data_failure attribute on the result object from a call. I still need to update the docs and tests etc.

[vsajip]

I have now pushed the changes, and updated the documentation which you can see at https://gnupg.readthedocs.io/en/latest/ - search for "New in version 0.5.4" to see the update.

[martinv13]

Thanks!